GUIDE · RANSOMWARE RECOVERY

Backup restore testing: how to do it and how often

This guide explains how to plan and carry out a backup restore test, what to measure during the test and how to record the outcome. By the end you will know which systems to test first, how often to do it and who must approve the result. A backup that has never been restored remains an unverified promise.

Updated october 2026
10 min read
Altovar Technical Editorial Team

1. Why an untested backup is not protection

A backup is created by software that finishes without errors, but this says nothing about whether the data can be restored into a working system. Configuration errors, missing encryption keys, dependencies that were not saved or systems that fail to restart are common problems, and they almost always surface at the worst moment: during an incident. The restore test turns confidence into proof, with a measured and recorded result.

In the context of ransomware the test has an additional value. It shows whether the copies are genuinely usable after an attack, whether the return times are compatible with the work, and whether the restore credentials work even when the main network is unavailable. Without this evidence, any discussion of resilience remains theoretical and hard to defend with an insurer or an auditor.

2. What to test: scope and priorities

It is not realistic to test every system every time. The scope should be chosen according to criticality: which data or applications, if unavailable, would block the work or create obligations towards customers and authorities. Include in the scope at least one full restore case, meaning a system with its data and its dependencies, not just a single file, because a single file does not prove that the application will work again.

Not all tests have the same purpose. Recovering a single file checks the everyday procedure. Restoring a whole server checks timings and dependencies. A test in an isolated environment verifies that the copy is clean. Choose the type of test according to the question you want to answer, and document which question each test answers.

3. Prerequisites: environment, documentation and owners

Before starting you need a test environment separate from production, a copy of the documentation for the systems involved, and a test owner with sufficient authority to stop the test if it risks interfering with real work. The test environment must not be connected to the production network without controls, because the test could reintroduce malware or send data to real systems.

You also need to know in advance which credentials are required for the restore and where they are kept. Ask the provider that manages the backups to state who can authorise a restore and how to proceed if the main network is unavailable. This information must be available outside the system being tested, otherwise the test proves nothing about the real case.

PrerequisiteWho looks after itCheck
Separate test environmentIT manager or supplierNo direct connection to production
System documentationProcess ownerDependencies and start-up order listed
Restore credentialsOwner or delegateAccessible even without the main network
Test authorisationOwnerWritten approval with scope and date

4. How to run a restore test

The test should be run as a controlled trial, with a known sequence and a log of every step. Improvising during the test makes it hard to tell which part worked and which did not. The procedure below is designed for the test restore of a system with data and applications, to be adapted to your infrastructure. Note each step as you carry it out, because details slip away quickly and reconstruction afterwards is always less precise.

It is best to run the test at a quiet time, with the owner present and an observer recording the times. At the end, compare the result with the expected outcome and open a corrective action for every deviation. A failed test is a useful result, provided it produces a concrete correction. If the test takes longer than expected, that is not a failure in itself: it is a figure to measure and bring into the minutes.

  1. Choose the system and the reference copy, noting the date of the copy.
  2. Prepare the test environment and verify that it is not connected to production.
  3. Restore the copy into the test environment, recording the start time.
  4. Check that the system starts and that its dependencies are available.
  5. Verify data integrity with a sample of known files or records.
  6. Perform a typical piece of work and compare the result with the expected outcome.
  7. Record the end time, errors and decisions taken, then dismantle the test environment.

5. What to measure: time, integrity and completeness

The test must produce measurements, not impressions. The three essential dimensions are the time taken to return to a usable system, the completeness of the recovered data and integrity, meaning certainty that the data is what was expected and has not been altered. The measured time should be compared with what the business can tolerate, and the difference is the concrete measure of residual risk.

For each dimension define an acceptance criterion before the test, so that the result does not depend on the impression of whoever observes it. A criterion can be qualitative, such as all essential services starting, or tied to a time the organisation has decided to accept. Write the criterion in the test minutes before you begin. A criterion written before the test prevents the result from being judged after the fact, based on what seems acceptable.

  • Recovery time measured from the start to the verification of the service.
  • Recovered data compared with the reference list.
  • Samples of files opened and read without errors.
  • User access verified with normal credentials.
  • Any missing dependencies noted with the owner.
  • Outcome compared with the criterion written before the test.

6. Test frequency and what changes between tests

Frequency should be chosen according to the criticality of the systems and how quickly they change. There is no universal cadence: for critical systems a regular, planned test is advisable, and also each time an application, an infrastructure or the backup system changes significantly. A fixed cadence, approved by the owner and recorded in the calendar, prevents the test from slipping until it is too late to matter.

Something almost always changes between tests: software versions, configurations, the people who manage the system. For this reason the test is not an activity to repeat identically, but a check that updates the documentation. After each test, review the scope and ask whether it still covers the systems that matter most for today's work. Every change to the network or to the backup systems is an opportunity to ask whether the previous test still holds.

7. Common mistakes in testing

The first mistake is testing only the files that are easy to recover, leaving out complex systems and their dependencies. The second is treating as successful a restore that starts but is not usable, because an application's data or service credentials are missing. The third is running the test on the production network, risking the reintroduction of problems or the exposure of real data.

Another mistake is not documenting. A test without minutes proves nothing in front of an auditor or an insurer, and leaves no trace of the problems resolved. A lack of a clear owner is also common: if nobody is authorised to declare the outcome, the test becomes an activity carried out out of habit and then forgotten. A mistake that recurs across several tests is a sign that the procedure needs changing, not merely repeating more carefully.

8. Related questions to read next

This article focuses on proving backups. Some related questions deserve separate study, and it is best to tackle them in the order that fits your situation. Related topics include the operational guide to recovery after ransomware, building an isolated recovery environment, verifying that copies are not already compromised, rebuilding the corporate directory, the order in which to restart systems, offline solutions compared, and documenting tests for auditors and insurers.

If you have time for only one further read, choose first the verification of compromised backups, because a test that uses an infected copy proves nothing. Then move on to documenting tests, which you will need for every later check. The other reads become useful once the first test has shown where the gaps are. Each related deep dive answers a question that emerges from the test, so the test becomes the starting point for all of them.

9. How to start: the first test this week

The first practical step is to choose a single system, the one whose outage would block the most work, and ask your backup provider for the date of the last copy and the last test run on it. If no test exists, that is the first entry in the minutes. You do not need a perfect environment to begin: you need to know whether the copy is usable.

In the same week, set a meeting with the test owner and with whoever can approve the result. Define the acceptance criterion in two written lines and prepare a separate test environment, even a simple one. The real test can come later, but the decision to run it and the chosen scope must be written down today. The minutes can be simple, as long as they contain the date, the scope, the criterion and the outcome of each check made during the test.

Frequently Asked Questions

How often should a restore test be carried out?

There is no single cadence that suits everyone. A planned, regular test is advisable for critical systems, plus a test every time an application, an infrastructure or the backup system changes significantly. The frequency should be set by the owner according to the downtime the organisation can tolerate, and recorded in the calendar like any other deadline.

Is a successful test on a single file enough?

No. Restoring a single file shows that the everyday procedure works, but it does not tell you how long it takes to bring a full system back or whether the applications start up again with their data. A serious assessment requires at least a restore of a complete system, with its dependencies, in a separate environment.

Can I run the test on the production network?

It is risky and generally discouraged. A test restore can bring back malware, overwrite real data or send information to connected systems. The test environment must be separate and controlled. If you do not have one, the first useful investment is exactly this, even in a simple form, to be defined with your IT supplier.

Who must approve the outcome of the test?

The owner or a delegate with sufficient authority. The IT manager carries out the test and records the data, but the decision about what to accept as a result, and which corrections to make, is a governance activity. This separates those who do the work from those who answer for the risks, and it is something auditors and insurers look at closely.

Can the backup provider run the tests on my behalf?

They can carry them out, but responsibility for the result remains yours. Ask the provider to give you a written record with scope, criteria, times and outcome, and to state any limits they encountered. Also check whether the test covers your real situation, because a generic test may not represent your case.

What happens if the test fails?

A failed test is valuable information, not a catastrophe, if it happens during a planned trial. Record the point of failure, identify the cause, correct the procedure or configuration and repeat the test on that point. The real risk is not knowing, and discovering it only when a real restore is needed.

About this content

General informational guide. Regulations and technical requirements change over time: always check the official sources and assess your specific case with a qualified technical or legal advisor.

GET IN TOUCH

Talk to our team.

We'll show how Altovar integrates into your infrastructure. No commitment, no sales pitch. Just a technical conversation.

BOOK A DEMO →